Protein-folding chaperones predict structure-function relationships and cancer risk in BRCA1 mutation carriers

Abstract

Predicting the risk of cancer mutations is critical for early detection and prevention, but differences in allelic severity of human carriers confound risk predictions. Here, we elucidate protein folding as a cellular mechanism driving differences in mutation severity of tumor suppressor BRCA1. Using a high-throughput protein-protein interaction assay, we show that protein-folding chaperone binding patterns predict the pathogenicity of variants in the BRCA1 C-terminal (BRCT) domain. HSP70 selectively binds 94% of pathogenic BRCA1-BRCT variants, most of which engage HSP70 more than HSP90. Remarkably, the magnitude of HSP70 binding linearly correlates with loss of folding and function. We identify a prevalent class of human hypomorphic BRCA1 variants that bind moderately to chaperones and retain partial folding and function. Furthermore, chaperone binding signifies greater mutation penetrance and earlier cancer onset in the clinic. Our findings demonstrate the utility of chaperones as quantitative cellular biosensors of variant folding, phenotypic severity, and cancer risk.

Keywords: BRCA1; CP: Cancer; HSP70; HSP90; cancer; expressivity; genetic variation; hypomorphism; penetrance; protein-folding chaperones; variants of uncertain significance.

Figure 1.. Most pathogenic BRCA1-BRCT variants bind chaperones

(A and B) BRCA1 mutations cataloged in ClinVar (accessed April 27, 2023). (A) Clinical significance of BRCA1 mutations grouped by mutation type. (B) Location of pathogenic missense mutations cataloged in ClinVar with at least a “two-gold-star” review status (n = 285). (C) C-terminal BRCA1 3×FLAG-tagged truncation construct used for LUMIER. The two BRCT subdomains are colored (orange and wheat). pSXXF denotes the BRCT binding phosphopeptide (BACH1). Protein Data Bank (PDB): 1T29. (D) LUMIER with BACON (bait control) approach to quantify chaperone interactions in HEK293T cells. Diagram was created with BioRender. (E) Chaperone binding to benign and pathogenic BRCA1-BRCT variants in ClinVar (including “likely” classifications). Z scores were calculated by averaging non-transfected wells to illustrate the significance of the data. Dashed line indicates the threshold for a statistically significant signal (Z score > 2.5). Wild-type BRCA1-BRCT values are shown as green-filled triangles. Chaps, chaperones; Patho., pathogenic. (F) BRCA1-BRCT variant levels after pull-down detected by ELISA. (G) Variant ΔΔG values (kcal/mol) predicted by FoldX relative to the wild-type value (ΔΔG = 0). Dashed line indicates the threshold for structure disruption (ΔΔG > 2). (H) HSP70/HSP90 interaction preferences. Diagonal shows the identity line. The unknown group includes variants annotated as having uncertain significance, no significance provided, conflicting interpretations, or “one-gold-star” review status. FANCA variants and a frameshifted (fs) variant encoding an additional out-of-frame HSP70 site shown for comparison. Statistical significance was determined using two-tailed Mann-Whitney t test (E–G). ****p ≤ 0.0001. Data are presented as mean values from at least two independent experiments.

Figure 2.. Chaperone biosensors detect a diversity of protein-folding variants

(A) Variants targeting secondary structure elements in the BRCA1-BRCT domain. (B) Y1845 variants that disrupt or support hydrophobic interactions. FoldX ΔΔG predictions for each variant are shown. (C) T1685 variants that disrupt or support side-chain hydrogen bonding. Two different T1685S codons were tested because these codons previously exhibited different functional effects. (D and E) Chaperone binding predictions to the BRCA1-BRCT variant library. The “disruptor” bin includes variants that introduce prolines in α-helices; truncations within the BRCT domain; buried variants that introduce a charge, decrease hydrophobicity, or create a steric clash; and charged variants that mutate residues in hydrophobic networks or disrupt side-chain hydrogen bonding. The “no effect” bin includes variants that fully delete the BRCT domain; variants outside the BRCT domain; buried variants that retain hydrophobicity; isosteric or quasi-isosteric variants; variants that maintain side-chain hydrogen bonds; and other surface variants not included prior. Dashed line indicates the cutoff for binding (HSP70 score > 0.5). (E) excluded variants previously characterized using protease sensitivity. Statistical significance was determined using a two-tailed Mann-Whitney t test. ****p ≤ 0.0001. Data are presented as mean ± standard deviation values from at least two independent experiments.

Figure 3.. The degree of chaperone binding is proportional to the severity of BRCA1 mutation

(A and B) Correlation of HSP70 binding with BRCA1-BRCT variant stability. Horizontal dashed line is the average of four double variants that combine strongly chaperone-bound single variants. Vertical dashed line reflects the upper limit of stability measurements owing to variant insolubility. The correlations were fit to a linear regression within the linear regime (FoldX: < 10 kcal/mol, empirical: < 5 kcal/mol). G1788V was not fit because the stability measurement for this variant was previously reported as contradictory. (C and D) HSP70 binding binned according to the functional effect class measured using aggregation/degradation or transcriptional activation assays. (E) Linear correlation of HSP70 binding and variant levels (ELISA) at different concentrations of cell lysates pulled down. Dashed lines indicate the average LUMIER and ELISA signals observed under our standard assay conditions for variants that bound strongly to HSP70. (F and G) BRCA1 variant function in HAP1 (cell fitness) or HeLa (HDR) cells binned according to the degree of HSP70 binding. (H) Integrated functional data from neXtProt binned by the magnitude of HSP70 binding. Percentages calculated using the percentage of variants associated with each HSP70 binding magnitude and phenotype intensity. Func., functional. Statistical significance was determined using Kruskal-Wallis ANOVA test (C, D, F, and G) or chi-squared test (H). ****p ≤ 0.0001, **p ≤ 0.01, and *p ≤ 0.05. Data are presented as mean ± standard deviation values from at least two independent experiments.

Figure 4.. Chaperones identify diverse classes of pathogenic BRCA1 variations in humans

(A) HSP70 interaction scores for natural human variants observed in patients with cancer (ClinVar and cBioPortal/TCGA^,) or the general population (gno-mAD). Unknown (unk.) includes variants annotated as having uncertain significance, no significance provided, conflicting interpretations, or one-gold-star review status. (B) The functional severity^, of natural BRCA1-BRCT variants grouped by the degree of HSP70 binding. LoF, loss of function. (C) HSP70 binding to natural BRCA1-BRCT human variants binned by the predicted effect on domain structure. Variants were ranked using an informed hierarchical approach (see STAR Methods). Moderate HSP70-bound variants are colored magenta and overlaid onto the BRCT crystal structure. An example long-range interaction that coordinates the two BRCT subdomains is indicated with an arrow. Struct., structure. (D and E) ROC curves using pathogenicity annotated in ClinVar (D) or mode phenotypic intensity annotated in neXtProt (E) as the target datasets. The neXtProt curves designate mild, moderate, and ambiguous (multimodal) variants as pathogenic. (F) The observed AUCs when mild, moderate, and ambiguous variants in the neXtProt target dataset were designated benign or pathogenic. Statistical significance was determined using Kruskal-Wallis ANOVA test (A and C) or chi-squared test (B). ****p ≤ 0.0001 and **p ≤ 0.01. ns, not significant. Data are presented as mean values from at least two independent experiments.

Figure 5.. Chaperone binding patterns stratify cancer risk and variant expressivity of human BRCA1 carriers

(A) BRCA1 mutation penetrance likelihood binned by the magnitude of HSP70 binding. Data shown were obtained from the Leiden Open Variation Database (LOVD) (accessed April 24, 2023). (B–D) Age of first cancer diagnosis for patients carrying BRCA1 mutations binned by mutation type (B), missense mutations in domains (C), and the degree of HSP70 binding (D). Statistical significance was determined using two-tailed (A) or one-tailed (B–D) Mann-Whitney t test. ****p ≤ 0.0001, **p ≤ 0.01, and *p ≤ 0.05. The number of variants in each bin is shown in parentheses. Data are presented as mean values from at least two independent experiments.